Change Summary for MPICH2 for week of Nov 13, 2005

Information on all weekly change summaries is available.

List of revisions

• (ashton) added Java requirement
• (thakur) added MPIR_ERRTEST_ALIAS_COLL that checks for buffer aliasing without MPI_BOTTOM exception for reduce/allreduce
• (ashton) added SIGCHLD signal handler for Macs since stdout and stderr don't close when a process exits
• (gropp) Added better error messages if unable to determine the lenght of Fortran integer and double precision quantities
• (thakur) added buffer alias error check
• (ashton) added comment
• (gropp) Added comments about the struct align; added a few cases. There are probably still problems with this
• (ashton) added error checking
• (ashton) added error handling comments
• (gropp) Added fixme about modifying mpif.h for Fortran 90
• (gropp) Added missing AC_LANG_PREFIX definition
• (gropp) Added simplified Fortran 90 handling to work around problems with the Fortran 90 extensions for autoconf in confdb
• (ashton) added zero copy shm code for the sshm and ssm channels. Only Windows uses the new code by default
• (rross) Adjustments to struct alignment tests to get us closer to right.
• (thakur) changed int * to MPI_Fint * in romio fortran interface
• (ashton) check for NULL after malloc
• (thakur) commented out concurrent_spawns test for now because it breaks mpd
• (ashton) corrected error name
• (gropp) Detect missing arguments
• (gropp) Ensure that there is a newline at the end of any message in mpiexec_usage
• (rross) even closer to right.
• (thakur) fixed bug
• (chan) Fixed documentation to get around the doctext bug related leading "'" characterin a line. Enhanced documentation.
• (ashton) fixed mpi export declaration when compiling the fortran interface dll
• (gropp) Fixed swapped indirect block size entries and added description of the indirect data structures
• (ashton) fixed the receiver's half of the zero copy shared memory code
• (ashton) fixed typo
• (rross) fixing 16-byte alignment test for ia32.
• (ashton) handle HAVE_MAX_INTEGER_ALIGNMENT, HAVE_MAX_FP_ALIGNMENT, HAVE_DOUBLE_POS_ALIGNMENT, HAVE_LLINT_POS_ALIGNMENT
• (ashton) ignore EOF errors and added some debugging printfs
• (lusk) included FAQ in Users Guide
• (chan) Matched the changes in mpe_log.c
• (rross) maybe this is it.
• (ashton) removed essm
• (chan) Replace the specific use of AC_TRY_LINK by AC_LINK_IFELSE + AC_LANG_PROGRAM to avoid the warning of "ignoring PROLOGUE".
• (thakur) return error if req create fails in mpid_recv_init
• (gropp) Switch from AC_TRY_COMPILE to AC_COMPILE_IFELSE to suppress Fortran missing-PROLOGUE warnings from autoconf
• (gropp) Test to see if gcc correctly handles -std=c98 (fails on Solaris with gcc 2.95.3)
• (ashton) typo
• (gropp) Updated CHANGES and RELEASE_NOTES for 1.0.3rc1
• (gropp) Updated common mpiexec documentation; includes a guess at commonly implemented extensions
• (gropp) Updated configure
• (buntinas) updated gasnet channel to work with new ch3 init mechanism
• (ashton) updated version

Files for each revision